This study investigates how methodological aspects of seed-bank sampling affect seed-bank records in temperate deciduous forests. We focused explicitly on seed-bank records of ancient forest species, which are assumed to lack a persistent seed bank; a hypothesis suspected to be partly due to methodological shortcomings. Through a quantitative review of 31 seed-bank studies in temperate deciduous forests of central and north-west Europe, we quantified the role of sampling methodology in constraining total seed-bank records and seed-bank records of ancient forest species (γ-diversity, average species' retrieval frequency and average seed density). A major methodological trade-off was established between sampled plot area and the number of plots: at an increased number of plots, the area sampled per plot decreased significantly. The total surface area sampled in a study was the primary determinant of γ-diversity, both for overall species richness and for ancient forest species richness. A high retrieval frequency of ancient forest species indicated that few plots were intensively sampled. The parallel increase in total species richness and ancient forest species richness and the non-significance of their ratio in relation to methodological variables suggests that ancient forest species are not particularly rare in the seed bank compared to other species. These results imply that sampling methodology has a far-reaching impact on seed-bank records such as γ-diversity, the detection of ancient forest species and ultimately seed-bank composition. We formulate a set of guidelines to improve the quality of seed-bank studies in temperate deciduous forests.